# Content Transfer Tool, Bulk Import and user migration

> **Verified:** 2026-09 against https://experienceleague.adobe.com/en/docs/experience-manager-cloud-service/content/migration-journey/cloud-migration/content-transfer-tool/guidelines-best-practices-content-transfer-tool and .../content-transfer-tool/user-mapping-and-migration

Open this when planning the content move, sizing the freeze window, or deciding between CTT and Bulk Import.

## Content Transfer Tool

1. Install CTT on the **source author** instance; check the disk, Java and AEM version prerequisites on the guidelines page.
2. Create a **migration set** in Cloud Acceleration Manager and copy its extraction key to the source.
3. **Extraction**: source → Adobe-managed staging storage. For large blob stores, use the pre-copy option so the binaries move ahead of the freeze.
4. **Ingestion**: staging → target environment. The first ingestion runs with **wipe**; later **top-ups** run without it and carry the delta since the last extraction.
5. Watch the logs in CAM, re-run failed steps, and validate counts (pages, assets, tags, users) against the source.

### What top-ups do and do not carry

Adobe's current guidelines state that **an asset updated or deleted in the source is respected during top-up operations** — deletes propagate. What still deserves verification on every rehearsal is **moves and renames**, and any content whose path changed as part of the refactor. Budget a validation pass that compares path counts per tree, not a spot check.

### Rules that save a weekend

- Purge versions, audit logs and workflow instances on the source before the first extraction. Transfer time scales with repository size, and most 6.x repositories are half history.
- Migrate **author**; publish is repopulated by publishing from the migrated author, which is also a free end-to-end test of the new publishing chain.
- An ingestion with wipe blocks the target environment while it runs. Schedule it, and never run one against production after go-live.
- `Handling Large Content Repositories` (Adobe's own guidance page) is worth reading in full before committing to a freeze window above a terabyte.

## Users, groups and identity

- Users and groups referenced by migrated content come across with CTT; there is a separate **Group Migration** step for groups that are not referenced.
- **Passwords are not migrated.** Authors sign in through Adobe IMS afterwards, so the real work is Admin Console: create product profiles, map them to AEM groups, and decide on MFA and directory sync (or bulk CSV upload) before UAT — not after.
- The `user-mapping-and-migration` documentation page describes how existing users are matched to IMS identities. Do the mapping before UAT so testers exercise real roles and real permissions; a UAT run as an administrator proves nothing about the permission model.

## Bulk Import for large asset libraries

Bulk Import pulls assets directly from cloud storage into AEM without streaming binaries through the JVM. Supported sources are **Azure Blob Storage, Amazon S3, Google Cloud Storage, Dropbox and OneDrive**. OneDrive uses delegated (user-based) OAuth only, so unattended imports should use Azure Blob or S3.

| | Content Transfer Tool | Bulk Import |
|---|---|---|
| Keeps paths, references and versions | Yes | No — assets land fresh |
| Metadata | Comes with the asset | Often needs a separate CSV import |
| Processing | Renditions regenerate on the target | Renditions generate on ingest |
| Best for | The site tree and the assets it references | A large standalone library, or an ongoing feed from a DAM |

Either way, every asset is reprocessed by asset microservices on arrival. Confirm the **processing profiles** are configured on the target *before* ingesting, or you will reprocess the whole library a second time.

## Cutover runbook shape

```
T-6w  Rehearsal 1: full extraction + ingestion to stage; time every step
T-3w  Rehearsal 2 at production volume; UAT against migrated content
T-1w  Code freeze on the source; final code deployed to the target
T-2d  Content freeze begins (authors paused; keep a change log for anything urgent)
T-1d  Final top-up; activate the publish tree; warm the cache
T-0   CDN/DNS switch; smoke tests; watch error rates and 404s
T+2w  Hypercare; source kept read-only as the fallback, then decommissioned
```

Cutover items that are forgotten until they break: redirects and vanity URLs, sitemap and `robots.txt`, form endpoints, integration callback URLs, scheduled jobs that must not double-run across both systems, and search engine verification records. Each one gets an owner and a verification step in the runbook.

## Freeze enforcement

A freeze that relies on an email to the authoring team is not a freeze. Enforce it: remove write permissions for author groups for the window, or put the source author into a read-only state, and keep a written change log of anything that had to go in anyway so it can be re-applied on the target by hand.
